lable at ScienceDirect

Journal of Cleaner Production 65 (2014) 97e107

Contents lists avai

Journal of Cleaner Production

journal homepage: www.elsevier .com/locate/ jc lepro

Energy efficiency analysis on Chinese industrial sectors: an improvedSuper-SBM model with undesirable outputs

Hong Li a,*, Jin-feng Shi b

a School of Economics, Peking University, Beijing 100871, Chinab School of Management, Shanxi University, Taiyuan, Shanxi 030006, China

a r t i c l e i n f o

Article history:Received 28 April 2013Received in revised form3 September 2013Accepted 22 September 2013Available online 30 September 2013

Keywords:Energy efficiencyIndustrial sectorsUndesirable outputsSuper-SBM model

* Corresponding author. Tel.: þ86 10 6275 5658; faE-mail address: hlipku0201@gmail.com (H. Li).

0959-6526/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.jclepro.2013.09.035

a b s t r a c t

In this article we proposed an improved Super-SBM model dealing with undesirable outputs under theweak disposability assumption of undesirable outputs. Energy efficiencies of various industrial sectors inChina from 2001 to 2010 are measured based on this model, and the influencing factors for energyefficiency are explored by Tobit regression model. Empirical results show that, during “The EleventhFive-year Plan”, energy efficiency of each industrial sector and category has been improved to variousextents, but overall efficiency variations among industries have not taken on a convergence trend. Lightindustry has achieved the highest energy efficiency, followed by heavy industry; while the energy ef-ficiency of the latter has a faster growth rate compared with that of light industry; the gap betweenthese two industries’ energy efficiency has been reduced. Energy efficiency variation presents an obviousfeature of industrial economy transformation. The analysis of influencing factors show that enterprisescale, industry concentration, industrial property rights structure, and government regulation all affectenergy efficiency apparently, but their effects vary across industries. Lastly, based on research results,this paper gives some policy recommendations on improving energy efficiency of the industrial sectorsin China.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Industry dominates national economy, and industrialization isthe core and foundation of economic modernization. Since the re-form and opening-up policy went into force 30 years ago, indus-trialization has been advanced quickly in China and gained manygreat achievements. Thereinto, industrial product outputs rose toNO.1 of theworld already andmanufacture outputs account for 20%globally now, which makes China the biggest manufacturingcountry. Meanwhile, industry is the main source of energy andresource consumption and pollutants in the country, and impedessustainable development seriously. The Chinese government hasadopted such policies as upgrading industrial structure and accel-erating technical progress. To improve energy efficiency is consid-ered as the basic principle in realizing energy-saving and emission-cutting. Hence, measuring and analyzing the influencing factors ofindustrial energy efficiency is the foundation to effectively boost

x: þ86 10 6275 1460.

All rights reserved.

industrial energy efficiency and build an industrial system withlow-energy consumption, low-pollution and low-emission.

Energy is important in our economic life and related with manyaspects such as economy, financial markets and social stability(Hamilton, 1983; Lardic and Mignon, 2008; Song et al., 2012, 2013;Yang et al., 2013). Energy efficiency has been studied for a long time.Based on the classical definition given by World Energy Council in1995 andPatterson (1996): energyefficiencymeans using less energyto produce at least equal number of services or useful outputs. Af-terward, Hu and Wang (2006) proposed two methods to measureenergy efficiency based on the number of elements that impact en-ergy efficiency duringmanufacturing process, i.e. single factor energyefficiency and total factor energy efficiency. At present, indexdecomposition analysis is used to measure single factor energy effi-ciency, usually including Index decomposition method proposed bySun (1998) and Fisher Index decompositionmethodproposed byAnget al. (2004). As formeasuring total factor energy efficiency, methodsmostly used include parametric method based on stochastic frontieranalysis (SFA) and non-parametric method based on data envelopanalysis (DEA). Numerous researchers have been studying industrialenergy efficiency. For instance, Jenne and Cattell (1983) examined the

1 In accordance with the global environmental conservation awareness, unde-sirable outputs of productions and social activities, e.g., air pollutants and haz-ardous wastes, are being increasingly recognized as dangerous and undesirable.Thus, development of technologies with less undesirable outputs is an importantsubject of concern in every area of production. Traditional efficiency measurementmodel usually assumes that producing more outputs relative to less input resourcesis a criterion of efficiency. In the presence of undesirable outputs, however, tech-nologies with more good (desirable) outputs and less bad (undesirable) outputsrelative to less input resources should be recognized as efficient.

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e10798

change in the ratio of energy consumed to industrial production from1968 to 1980, and described in detail the implications of the two re-cessions for efficiency and industrial structure. Phylipsen et al. (1997)identified structural differences in energy intensive industries andanalyzed the ways to incorporate these differences in internationalcomparisons of energy efficiency. They concluded that structuraldifferences could be taken into account in cross-country comparisonsof energy efficiency with appropriate physical energy efficiency in-dicators. DeCanio (1993) studied the relationship between energyefficiency and industrial investment and found that many in-vestments in energy efficiency fail to bemade despite their apparentprofitability. Bunse et al. (2011) presented a range of methods formeasuring and evaluating energy efficiency improvements inmanufacturing processes such as Key Performance Indicators (KPI)and Balanced Scorecard (BSC). Oda et al. (2007) compared energyefficiency across countries in power, steel, and cement sectors, andobtained the findings that new installation and continual mainte-nance were essential for energy efficiency.

In recent years, China’s industrial energy efficiency has becomea hot research topic. Andrews-Speed (2009) examined China’svigorous programs in reversing the trend of soaring national energyintensity and reducing the intensity by 20% over the period 2006e2010, and thus to evaluate the likelihood that policies of today willyield improvements over a longer period. It is suggested that Chinashould make greater efforts to address a number of deficiencies inthe following aspects: the reluctance to use economic and financialinstruments; the dependency of energy policy on industrial andsocial policies; the nature of political decision-making and of publicadministration; a shortage of skills and social attitudes to energy.Chen and Yeh (1998) employed data envelopment analysis tocompare energy utilization efficiency between mainland China andTaiwan from 2002 to 2007, and they found that Taiwan is better interms of energy utilization efficiency and the Eastern region ofChina had higher efficiency than the Western region. Zhou et al.(2010) summarized and analyzed the energy efficiency policies invarious industries from 1970 to 2010, and provided an assessmentof these policies and programs for understanding issues that willplay a critical role in China’s energy and economic future.Hasanbeigi et al. (2010) surveyed 16 cement plants with NewSuspension Preheater and pre-calciner (NSP) kiln and comparedthe plant energy consumption of both domestic (Chinese) and in-ternational best practice. They used the Benchmarking and EnergySaving Tool for Cement (BEST-Cement) and the bottom-up Elec-tricity Conservation Supply Curve (ECSC) model to estimate thepotential for the 16 studied cement plants in 2008. Hu (2012)examined the origin and development of energy conservationassessment (ECA) in China, and found that ECA has a great potentialin energy efficiency improvement and GHGs reduction. Wang et al.(2012) evaluated energy efficiency by DEA model under theframework of total factor energy efficiency for industrial sectors ofChina. They found regional energy efficiency disparity in China isprominent because of technological differences, and large scaleinvestment should be suspended in the country, especially incentral and western regions.

From social, economical and biological perspectives, under-performing industrial sectors are actually gambling the odds in apublic goods game. Neglecting efficiency, the industrial sectors gainsome profit while in terms of the public goods in China, and there isa loss derived from individual incentives (of the firms in a givensector) of not using more efficient procedures. Many theoreticalmodels on the public goods game have been employed to addresssuch dilemmas in an agent-principal setting and explore the factorsthat contribute to such an unfavorable outcome (see Perc andSzolnoki (2010) and Perc et al. (2013) for a comprehensive under-standing on coevolutionary games and group interactions). In this

paper, in order to fully reflect the situations and influencing factorsof industrial energy efficiency in China, we propose an improvedSuper-SBM model dealing with undesirable outputs1 under theweak disposability assumption of undesirable outputs. By this ef-ficiency model, we will measure energy efficiency of the industrialsectors from 2001 to 2010 and explore the influencing factors ofenergy efficiency.

The rest of the paper is organized as follows. In Section 2 wepropose the improved Supper-SBMmodel dealing with undesirableoutputs under theweak disposability assumption, and illustrate theTobit regression method. In Section 3 we present the industrialsectors and its categories, inputeoutput data and the analysis re-sults of energy efficiency in Chinese industrial sectors from 2001 to2010. We explore influencing factors of the energy efficiency byTobit regression model in Section 4 and give the conclusions inSection 5.

2. Methodology

2.1. Improved SBM and Super-SBM models

Tone (2001) proposed a slacks-based measure (SBM) of effi-ciency. Unlike traditional DEA model, the slack variables in SBMmodel are directly added into the target function. The SBMmethodis thus non-radial and deals with input/output slacks directly,eliminating the radial and oriented deviation (Song et al., 2013). Inaddition, the economic interpretation of the evaluation model is tomake actual profit maximization rather than simply maximize ef-ficiency ratio as in CCR or BCC model. As known to all, undesirableoutputs such as waste water, exhausted gas, are unavoidable forproduction and living, so it is necessary to take account the unde-sirable outputs into efficiency evaluation model (Seiford and Zhu,2002). Tone and Sahoo (2003) proposed a new measuring effi-ciency scheme with undesirable outputs based on SBM model. Inorder to better handle the undesirable outputs, we should catego-rize disposability assumptions, that is, strong disposability or weakdisposability. As pointed out in Färe and Grosskopf (2004) and Ray(2004), the strong disposability states that undesirable outputs canbe discarded at no cost which undermines the usefulness of theconcept; while weak disposability assumes that the outputs areweakly disposable while only the sub-vector of the desirable out-puts is strongly disposable. In this paper, we will combine theassumption of weak disposability with the SBM model as well asSuper-SBM model, and thus obtain an improved efficiency modelconsidering the undesirable outputs.

Now we consider a production system with n DMUs, each unithas three factors: inputs, desirable outputs and undesirable outputs(environment pollution, such as CO2, SO2, etc), as represented bythree vectors: x˛Rm; yg˛Rs1 ; yb˛Rs2 . We define the matrices X, Yg, Yb

as follows:

X ¼ ½x1; x2;.; xn�˛Rm�n;Yg ¼ �yg1; y

g2;/; ygn

�˛Rs1�n;Yb

¼hyb1; y

b2;.; ybn

i˛Rs2�n;

Table 1Names and details of 36 sub-industries.

Industry classification Industrycode

Industry name

Mining industry (I) SER 01 Coal mining and washingSER 02 Oil and natural gas miningSER 03 Ferrous metal miningSER 04 Non-ferrous metal miningSER 05 Non-metal mining

Light industry (II) SER 06 Agricultural products processingSER 07 Food manufacturingSER 08 Beverage manufacturingSER 09 Tobacco manufacturingSER 10 Textile industrySER 11 Textile clothes, shoes, hats manufacturingSER 12 Leather, fur, feather manufacturingSER 13 Wood processing, and wood, bamboo, cane,

palm, and straw manufacturingSER 14 Furniture manufacturingSER 15 Papermaking and paper productsSER 16 Press and intermediary replicationSER 17 Cultural, educational and sports goods

manufacturingHeavy industry (III) SER 18 Oil processing, coking and nuclear fuels

processingSER 19 Manufacturing of chemical materials and

productsSER 20 Manufacturing of medicinesSER 21 Manufacturing of chemical fiberSER 22 Manufacturing of rubberSER 23 Manufacturing of plasticsSER 24 Manufacturing of non-metal productsSER 25 Smelting and rolling process of non-

ferrous metalSER 26 Smelting and rolling process of Ferrous

metalSER 27 Manufacturing of metal productsSER 28 Manufacturing of ordinary machinerySER 29 Manufacturing of special equipmentsSER 30 Manufacturing of transportation and

equipmentsSER 31 Manufacturing of electric machinesSER 32 Manufacturing of communication device,

computers and other electronicequipments

SER 33 Manufacturing of instruments, cultural andofficial mechanics

Electricity, gas andwater industry(IV)

SER 34 Production and supply of electricity, powerSER 35 Gas production and supplySER 36 Water production and supply

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e107 99

and assume that X > 0, Yg > 0, Yb > 0. Then the production possi-bility set (P) is defined by assuming

P ¼n�

x; yg ; yb����x � Xl; yg � Ygl; yb ¼ Ybl; l � 0

o;

where l is the intensity vector. According to the SBMmodel in Tone(2003) and weak disposability assumption, the improved SBMmodel dealing with undesirable outputs for evaluating DMUðx0; yg0; yb0Þ is as follows.

½Improved SBM� r* ¼ min1� 1

m

Pm

i¼1

S�i

xi0

1þ 1s1

Ps1r¼1

Sgrygr0

Subject to

x0 ¼ Xlþ S�yg0 ¼ Ygl� Sg

yb0 ¼ YblS� � 0; Sg � 0; l � 0

(1)

where S ¼ (S�, Sg) corresponds to the slacks in inputs and desirableoutputs. The optimization function value of r* is the efficiencyvalue of the Decision Making Unit ðx0; yg0; yb0Þ. m, s1 and s2 stand forthe number of factors for inputs, desirable outputs and undesirableoutputs. By CharneseCooper transformation, we can transform theabove nonlinear program into a linear program in the followingequivalent form.

½Improved SBM0� s* ¼ mint � 1m

Pmi¼1

S�ixi0

Subject to

1 ¼ t þ 1s1

Ps1r¼1

Sgrygr0

x0t ¼ XLþ S�

yg0t ¼ YgL� Sg

yb0t ¼ YbL

S� � 0; Sg � 0;L � 0; t > 0

(2)

However, most empirical results of the efficiency evaluationresearch have a common phenomenon, that is, plural DecisionMaking Units have the “efficient status” denoted by 100%. So howto rationally discriminate between these efficient DMUs isimportant for efficiency ranking and influence factors analysis.Based on the above improved SBM model, we improve the Super-SBM model in Tone (2002) under the assumption of weakdisposability. The improved Super-SBM model dealing with un-desirable outputs which is used for evaluating the SBM-efficientDMUs is as follows:

½Improved Super� SBM� d* ¼ min1m

Pm

i¼ 1

xixi0

1s1

Ps1r¼ 1

ygrygr0

Subject to

x � Pnj¼1;s0

ljxj

yg � Pnj¼1;s0

ljygj

yb ¼ Pnj¼1;s0

ljybj

x � x0; yg � yg0; y

b � yb0; yg � 0; l � 0

(3)

It is worth noting that the above improved SBM model and theSuper-SBM model dealing with undesirable outputs are under theassumption of constant returns-to-scale (CRS). Also we also canrelax and extend these models to the variable returns-to-scale(VRS) case with the restrictions

Pni¼1 li ¼ 1 in model (1) andPn

i¼1;s0 li ¼ 1 in model (3) respectively.

2.2. Tobit regression model

Although the improved efficiencymodel dealingwith undesirableoutputs can contribute to the improvement of the efficiency perfor-mance, it does not capture the key factors affecting the energy

Table 2Descriptive statistical characteristics of input and output variables.

Variable Inputs Outputs

Capitalx1 Laborx2 Energy x3 Industrial output yg Wastes yb

Mean 3.0825 1.8707 4.4717 9.0646 7.4405Median 1.4892 1.2415 1.3226 4.8643 0.3190Maximum 47.9014 7.7275 56.4130 55.4526 1291.2890Minimum 0.1110 0.1400 0.0955 0.1849 0.0033Std. Dev 5.1395 1.6177 8.3408 10.6865 72.8702Skewness 5.0705 1.1391 3.4669 1.9756 15.8621Kurtosis 35.9951 3.5349 17.2509 6.7711 272.3866

Note: Each sample has 360 observations, for the panel data includes 5 indicators of36 industrial sectors from 2001 to 2010.

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e107100

efficiency. Nevertheless,we couldusemultivariate analysis to explorethe influencing factors. Because the efficiency value is non-negative,the general estimation method as Ordinary Least Square (OLS) willlead to biased and inconsistent estimating results. Tobit regressionmodel is based on the principle ofmaximum likelihood estimation toget the consistent parameter estimation. This regression model be-longs to the limited dependent variable or truncation econometricsmodel, and the dependent variable can only be observed in arestricted way. The standard Tobit regression model is as follows:

y*i ¼ bXi þ miwN�0; s2

�; i ¼ 1;2;.;n; (4)

where i stands for the ith DMU, y*i a latent (i.e. unobservable)variable, Xi is K � 1 matrix on behalf of independent variables.m is stochastic error and submits to N(0,s2). The limited valueyi is:

yi ¼(y*i ; y*i > 00; y*i � 0

:

3. Data and empirical results

3.1. Data and industrial sectors

Currently, existing researches focus on energy efficiency of onesingle sector and disregard the undesirable outputs such as envi-ronment pollution. This paper is more concerned with distributioncharacteristics and the evolving trend of energy efficiency in thewhole industry of China. Hence, we select 36 sub-industries inChina from 2001 to 2010 and classify them into 4 categories, i.e.mining industry, light industry, heavy industry, and electricity, gasand water industry (see Table 1). Subsequently, we can not onlyanalyze the characteristics and disparities of energy efficiencyacross different sub-industries, but also study the energy effi-ciencies, evolving trends and influencing factors of different in-dustrial categories. In addition, the selected period spans two “Five-Year Plans”,2 i.e. “The Fifth Five-Year Plan” and “The Eleventh Five-Year Plan”. So we can have a deeper understanding of the variationof energy efficiency of different industries during economic growthand the relationship between the formation and implementation ofenergy-saving and emission-reducing polices.

In researches on energy efficiency, input indicators generallyinclude capital, labor, and energy consumption. In this paper, weutilize the outstanding net value of fixed asset of the enterprisesabove designated scale as the proxy for capital input; averageamounts of total employees of the enterprises above designatedscale as the proxy for labor input; total energy consumption of sub-industries as the proxy for energy input. Basically, it is recognizedthat DMU is relatively more effective duringmanufacturing processwhen the ratio of input over output is as little as possible. However,besides the desirable outputs, undesirable outputs such as envi-ronment pollutants are generated during manufacturing process aswell. Hence, we divide outputs into desirable outputs and

2 The five-year plans of People’s Republic of China (PRC) are a series of social andeconomic development initiatives. The economy was shaped by the CommunistParty of China (CPC) through the plenary sessions of the Central Committee andnational congresses. The party plays a leading role in establishing the foundationsand principles of Chinese communism, mapping strategies for economic develop-ment, setting growth targets, and launching reforms. Planning is a key character-istic of centralized, communist economies, and one plan established for the entirecountry normally contains detailed economic development guidelines for all itsregions. In order to more accurately reflect China’s transition from a Soviet-styleplanned economy to a socialist market economy (socialism with Chinese charac-teristics), the name of the 11th five-year program was changed to “guideline”.

undesirable outputs. We define the gross output value of the en-terprises above designated scale as desirable output, and we proxythe total emission of the three wastes (waste gas, waste water, in-dustrial residue) in all sub-industries for undesirable outputbecause it lacks accurate data of pollutants emission like CO2, SO2,NO2 in China. Data is cited from China Statistical Yearbook andChina Energy Statistical Yearbook published by the National Bureauof Statistics of China. The descriptive statistical characteristics ofabove mentioned input/output variables are shown in Table 2.

Table 2 suggests that median of different indicators is muchsmaller than themean value, and a larger standard deviation showsunbalanced production status of different industrial sectors, whichis more prominent in terms of desirable outputs and undesirableoutputs. Moreover, the maximum can be as much as more than 500times of the minimum for energy consumption and outputs indifferent industrial sectors and both are closely related to industryattributes. Therefore, an analysis in-depth of the energy efficiencyin industrial sectors of China and the exploration of influencingfactors are helpful for the implementation of energy conservationand emission reduction policies during the “Twelfth Five-YearsPlan”. Although not all industrial sectors are energy-intensive, wefind that correlation between energy consumption and desirableoutputs is still up to 0.5604 from the Pearson coefficient in Table 3.In other words, the DMUs’ production process has some so-called“isotonicity”. In addition, the correlation between inputs and un-desirable outputs is weak and insignificant, in line with the actualproduction expectation. Therefore, the energy efficiency measuredby the improved Super-SBM model is reliable, and the researchresults are completely believable.

3.2. Energy efficiency performance in China’s industrial sectors

In this section, we will measure the energy efficiency of Chineseindustrial sectors from 2001 to 2010 by the improved Super-SBMmodel, and analyze the energy efficiency performance and devel-opment trends in different industrial sectors.

3.2.1. Energy efficiency characteristics of Chinese industrial sectorsin 2001e2010

Fig.1 presents that Tobaccomanufacturing (SER 09), Leather, fur,feather, manufacturing (SER 12) and Manufacturing of electricmachines (SER 31) share the highest energy efficiency, but theenergy efficiency in Manufacturing of electric machines (SER 31) isabove 0.5 almost every year. The energy efficiencies of agriculturalproducts processing (SER 06), agricultural products processing (SER11), furniture manufacturing (SER 14), oil processing, coking andnuclear fuels processing (SER 18), communication device, com-puters and other electronic equipments (SER 32) andmanufacturing of instruments, cultural and official mechanics (SER33) are higher than the average efficiency value. However, coalmining and washing (SER 01), oil and natural gas mining (SER 02),

Table 3Pearson coefficients between input and output variables.

Indicator x1 x2 x3 Indicator x1 x2 x3

yg 0.6295*** (12.4851) 0.7453*** (21.1482) 0.5604*** (12.8038) yb 0.1361*** (2.5991) 0.0438 (0.8301) 0.1020* (0.0532)

Note:“*”,“**”,“***” present their significance respectively at levels of 10%, 5% and 1%.

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e107 101

production and supply of electricity, power (SER 34), gas produc-tion and supply (SER 35) and water production and supply (SER 36)have the lowest energy efficiency and high energy consumption. Onthe whole, energy efficiencies of most industrial sectors are rela-tively lower, including and efficiency values ranges from 10% to 50%,which suggests a big gap compared with western developedcountries.

Fig. 2 shows that various sectors (except water production andsupply) of mining industry (I) and electricity, gas and water in-dustry (IV) present the rising energy efficiency during 2001e2010,and efficiency disparity among sectors has been not significantlydecreased. Non-ferrous metal mining (SER 04) and ferrous metalmining (SER 03) of mining industry (I) have the highest energyefficiency, whereas coal mining and washing (SER 01) gain thelowest efficiency. Production and supply of electricity, power (SER34), gas production and supply (SER 35) of electricity, gas andwaterindustry (IV) almost share the same level of energy efficiency, butthe energy efficiency of water production and supply (SER 36) isrelatively low and shows no variation during these years.

Fig. 3 presents that energy efficiencies of various sectors in lightindustry (II) have increased during 2001e2010 and efficiencygrowth rates remain quite stable. Tobacco manufacturing (SER 09)and Leather, fur, feather, manufacturing (SER 12) have relativelyhigher energy efficiency, and energy efficiencies of textile industry(SER 10), papermaking and paper products (SER 15) and press andintermediary replication (SER 16) are lower. From Fig. 4, we learnthat except for manufacturing of non-metal products (SER 24),energy efficiencies of various sectors in heavy industry (III) have thesame variation tendency as those in other industrial categories. Inaddition, oil processing, coking and nuclear fuels processing (SER18) and manufacturing of electric machines (SER 31) have higherenergy efficiency than that of other sectors in light industry, whilemanufacturing of non-metal products (SER 24) has lower energyefficiency.

3.2.2. Energy efficiency of the four major industrial categories andtheir tendency analysis

According to the energy efficiency of each industrial sector andindustry classification in Table 1, we can give the efficiency of the

0.0

0.4

0.8

1.2

1.6

2.0

SER

01

SER

03

SER

05

SER

07

SER

09

SER

11

SER

13

SER

15

SER

17

SER

19

SER

21

SER

23

SER

25

SER

27

SER

29

SER

31

SER

33

SER

35

Fig. 1. Boxplot of energy efficiency of industrial sectors in 2001e2010 in China.

four major industrial categories in China from 2001 to 2010 andtheir evolving tendency in Table 4 and Fig. 5.

Table 4 shows that, the annual average energy efficiency valuesof mining industry(I), light industry(II), heavy industry (III) andelectricity, gas and water industry (IV) during 2001e2010 are0.1928, 0.3744, 0.3584, 0.1674 respectively. It indicates that underthe current conditions, energy efficiency of mining, electricity, gasandwater industry can be upgraded by at least 15% to catch upwiththe annual average energy efficiency of light and heavy industry.Fig. 5 shows that during 2001e2010, light industry obtains thehighest energy efficiency among the 4 industrial categories, fol-lowed by heavy industry. Energy efficiency of heavy industry isimproved more than the light industry, and the energy efficiencygap between them is narrowed due to multiple factors such asindustry attributes, energy structures and policies orientations. Inreal manufacturing process, light industry depends less on re-sources and develops intensively. Meanwhile, the technology andequipment has approached closely to world leading level. Whileheavy and mining industries as energy-intensive industries, mainlyrely on infrastructure resources like coals, and have extensivedevelopment mode and suffer laggard technology equipment.During “The Eleventh Five-Year Plan”, the government proposedpolicies to transform the economic development structure as wellas to cut down energy consumption and reduce emissions. Forexample, the related policies included integrating the mining in-dustry of Shanxi; sweeping out laggard production capacity of ironand steel industry; shutting down enterprises producing largeamount of pollution. Hence, energy efficiencies of different in-dustries in China have been promoted to various degrees. Heavyindustry has transformed from extensive pattern to intensivepattern, and attained higher energy utilization efficiency above 0.5in 2010. By contrast, energy efficiencies of mining industry andelectricity, gas, and water industry are relatively lower, and achieveno improvement compared with light and heavy industry. Finally,the energy efficiency of the entire industry in China is not satis-factorily high, which needs a further improvement.

3.2.3. Comparison of energy efficiency across industrial sectors inthe two “Five Year Plans”

“The Five Year Plan” is an important component of the long-term development in terms of the national economy, and itmainly concentrates on the planning of national construction,distribution of productivity and the proportion of national econ-omy. Therefore, wewill use the developing programs of “The Tenth-Five Plan” and “The Eleventh-Five Plan” to make further analysis onthe energy efficiency characteristics in China. The energy efficiencydifferences in different industrial sectors between the two periodscan be discerned in Fig. 6.

Fig. 6 illustrates that during “The Tenth-Five Plan” and “TheEleventh-Five Plan”, except for manufacturing of non-metal prod-ucts (SER 24) and manufacturing of electric machines (SER 31),every industrial sector has shown a significant improvement inenergy efficiency. This attributes to the fact that “The Tenth-FivePlan” gave priority to development and preliminarily establisheda market-oriented economic system, and the transformation ofeconomic growth pattern was rather low with a growing need ofenergy consumption. However, “The Eleventh-Five Plan” clearlystipulated that energy consumption per GDP should be lowered by

.04

.08

.12

.16

.20

.24

.28

.32

.36

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

SER01 SER02 SER03SER04 SER05

0.0

0.4

0.8

1.2

1.6

2.0

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

SER34 SER35 SER36

Fig. 2. Energy efficiency tendency of extractive industries and electricity, gas and water industries in China in 2001e2010.

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e107102

around 20%, and particularly incorporated energy-saving andconsumption reducing index into the five-year plan for the firsttime. “The Eleventh-Five Plan” required the implementation of thebasic national resources conservation and environment protectionpolicies. That means the establishment of the recyclable and sus-tainable national economic systemwith low input, high output, lowenergy consumption and low emissions, and the construction of aresource-saving and environment-friendly society.

On the whole, industrial energy efficiency variations in Chinaare characterized with three features: (a) term-end (2010) energyefficiency of 36 sub-industries is higher than that of term-

.2

.3

.4

.5

.6

.7

2002 2004 2006 2008 2010

SER06

.15

.20

.25

.30

.35

.40

.45

2002 2004 2006 2008 2010

SER07

.1

.2

.2

.3

.3

.4

.16

.20

.24

.28

.32

.36

2002 2004 2006 2008 2010

SER10

.35

.40

.45

.50

.55

.60

.65

2002 2004 2006 2008 2010

SER11

0

0

0

1

1

.2

.3

.4

.5

.6

.7

2002 2004 2006 2008 2010

SER14

.12

.16

.20

.24

.28

2002 2004 2006 2008 2010

SER15

.1

.1

.2

.2

.3

.3

Fig. 3. Energy efficiency tendency of ligh

beginning (2001). The entire industrial energy efficiency risesfrom 0.19 to 0.47, indicating an overall improvement of energyutilization efficiency. These results are in accordance with the tar-gets of “The Eleventh-Five Plan” to reinforce energy-saving andemission-cutting and reduce per GDP energy consumption. (b)Energy efficiency variations among industries have not shown anyobvious convergence characteristics. Though almost every indi-vidual industry has achieved a higher energy efficiency value, thestandard variation has been increased which might result fromindustry structure adjustment and strategy transformation. (c)Light industry has a better performance in the context of energy

5

0

5

0

5

0

2002 2004 2006 2008 2010

SER08

0.2

0.4

0.6

0.8

1.0

1.2

2002 2004 2006 2008 2010

SER09

.4

.6

.8

.0

.2

2002 2004 2006 2008 2010

SER12

.15

.20

.25

.30

.35

.40

.45

2002 2004 2006 2008 2010

SER13

0

5

0

5

0

5

2002 2004 2006 2008 2010

SER16

.2

.3

.4

.5

.6

.7

2002 2004 2006 2008 2010

SER17

t industries in 2001e2010 in China.

0.2

0.4

0.6

0.8

1.0

1.2

2002 2004 2006 2008 2010

SER18

.12

.16

.20

.24

.28

.32

2002 2004 2006 2008 2010

SER19

.15

.20

.25

.30

.35

.40

2002 2004 2006 2008 2010

SER20

.10

.15

.20

.25

.30

.35

.40

2002 2004 2006 2008 2010

SER21

.12

.16

.20

.24

.28

.32

2002 2004 2006 2008 2010

SER22

.2

.3

.4

.5

2002 2004 2006 2008 2010

SER23

0.0

0.4

0.8

1.2

1.6

2002 2004 2006 2008 2010

SER24

.10

.15

.20

.25

.30

.35

.40

2002 2004 2006 2008 2010

SER25

.1

.2

.3

.4

.5

2002 2004 2006 2008 2010

SER26

.20

.25

.30

.35

.40

.45

2002 2004 2006 2008 2010

SER27

.1

.2

.3

.4

.5

.6

2002 2004 2006 2008 2010

SER28

.1

.2

.3

.4

.5

2002 2004 2006 2008 2010

SER29

.2

.3

.4

.5

.6

2002 2004 2006 2008 2010

SER30

0.2

0.4

0.6

0.8

1.0

1.2

2002 2004 2006 2008 2010

SER31

.45

.50

.55

.60

.65

.70

2002 2004 2006 2008 2010

SER32

.2

.3

.4

.5

.6

2002 2004 2006 2008 2010

SER33

Fig. 4. Energy efficiency tendency of heavy industries in 2001e2010 in China.

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e107 103

efficiency, while heavy industry has a higher growth rate of energyefficiency. Nonetheless, the whole industry in China is still ofrelative low energy efficiency and demanding a furtherimprovement.

4. Influencing factors of energy efficiency in Chineseindustrial sectors

In the above study, we measure the energy efficiency of indus-trial sectors in China from 2001 to 2010, and analyze the high en-ergy efficiency disparity across various industries and someevolving trends. In order to further study the causes for the energyefficiency disparity, we will use the Tobit regression model toexplore relevant influencing factors on energy efficiency from such

Table 4Energy efficiency of Chinese four major industrial categories in 2001e2010.

Year 2001 2002 2003 2004 2005

I 0.0961 0.0936 0.1454 0.2060 0.1817II 0.2427 0.2500 0.2825 0.3743 0.3295III 0.2090 0.2186 0.3484 0.3301 0.3187IV 0.0487 0.0507 0.5993 0.0920 0.0943O 0.1912 0.1977 0.3192 0.3077 0.2846

Note: I-Extractive Industry, II-Light Industry, III-Heavy Industry, IV-Electricity, Gas and W

aspects as industry structure, energy consumption, technologyinnovation and government regulation. Meanwhile, wemake somepre-judgments for each factor.

4.1. Influencing factors’ selection

“The Tenth-Five Plan” and “The Eleventh-Five Plan” issued thehot topic that whether industrial economy transformation couldenhance of energy efficiency. Meanwhile, the industrial economy ofChina is basically in the stage of strengthening market competi-tiveness, boosting the reform of state-owned enterprises, andperfecting the legal structure of property rights. For this purpose,from the perspective of industry structure, we employ the industryscale and industry intensiveness as proxy for market structure,

2006 2007 2008 2009 2010

0.2060 0.2339 0.2540 0.2355 0.27560.3659 0.4233 0.4428 0.4715 0.56130.3584 0.4251 0.4395 0.4256 0.51100.1129 0.1340 0.1591 0.1708 0.21200.3193 0.3737 0.3915 0.3933 0.4701

ater Industry, O-Overall Average.

.0

.1

.2

.3

.4

.5

.6

.7

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Extractive IndustryLight IndustryHeavy IndustryElectricity, Gas and Warter IndustryOverall Average

Fig. 5. Energy efficiency of four major industrial categories in China from 2001 to 2010.

Table 5Influence factors and relevant symbols.

Variables Symbols Variables’ definitionand unit

Pre-judgment

Industrystructure

Enterprisescale

ES Ratio of gross outputvalue to enterpriseamount in sub-industry (0.1 billions/per unit)

Unknown

ES2 Quadratic of ES UnknownIndustryconcentration

IC Ratio of large andmedium sizedenterprises’ outputvalue to sub-industryoutput value (%)

Positive

Property-rightstructure

PS Ratio of state-ownedand state-ownedholding enterprises’output value tosub-industry outputvalue (%)

Negative

Capital-laborstructure

CL Ratio of net value offixed assets toemployees amount(10 thousands/perperson)

Unknown

Energyconsumption

Coal MC Ratio of coalconsumption tosub-industry’senergy consumption(%)

Negative

Oil OC Ratio of oilconsumption to sub-industry’s energyconsumption (%)

Negative

Electricity EC Ratio of electricityconsumption to sub-industry’s energyconsumption (%)

Positive

TechnologyInnovation

Research anddevelopmentinvestment

RDI Ratio of R&Dinvestment to sub-industry’s fixedassets (%)

Positive

R&D researchers Ln(RDR)

R&D researchers’logarithm

Positive

Governmentregulation

Pollution costs Ln(PC)

Logarithm of totalcosts of controlling(including variousindustries’ wastewater and waste gas)

Positive

Foreign domestic FDI Ratio of FDI and Unknown

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e107104

reflecting the industrial economies of scale to some extent; theproportion of state-owned and state-owned holding enterprises tothe gross industrial output value stands for property rights struc-ture, and capital-labor structure for human costs. In addition, weintroduce ES2 (see Table 5) into regression model in order toanalyze whether the variation of industry scale to energy efficiencyhas a U-type curve or inverse U-type curve. From the perspective ofenergy consumption, this paper also analyzes the impact of in-dustrial energy structure adjustment on energy efficiency in termsof coal, oil and electricity. Besides, research & development in-vestment and human capital are important support for technologyimprovement and innovation, so we utilize R&D investment (RDI)as proxy for the investment of searching technical progress, andR&D researchers (RDR) as proxy for the investment of R&D humancapital. In view of government regulations, we introduce foreigndirect investment as the indicator of the market openness ofdifferent industries and pollution control investment as the indi-cator of disposing costs resulted from governmental irrational be-haviors during marketization process. Given the huge diversity ofresources endowments, pollutant emissions, technology composi-tions and industrial cycle in China, we study the whole sample andindustry group sample for exploring the influencing factors of en-ergy efficiency on the level of the whole industry and differentindustries. The data is from China Statistical Yearbook, China En-ergy Statistical Yearbook and China environment Statistical Year-book. Table 5 shows the concrete definition of variable index andcorresponding symbols.

0.0

0.2

0.4

0.6

0.8

1.0

5 10 15 20 25 30 35

Tenth five year planEleventh five year plan

Fig. 6. Energy efficiency of industrial sectors in the two “Five Year Plans” in China.

investment investment fromHongkong, Macao andTaiwan to sub-industry’soutput value (%)

4.2. Tobit regression model of the energy efficiency and empiricalresults

Considering influencing factors, the Tobit regression model be-tween energy efficiency (EEI) in industrial sectors and influencingfactors is:

EEIit ¼ b0 þ b1ESit þ b2ES2it þ b3ICit þ b4PSit þ b5CLit þ b6MCit

þ b7OCit þ b8ECit þ b9RDIit þ b10lnðRDRÞitþ b11lnðPCÞit þ b12FDIit þ εit

(5)

EEIit on the left means the energy efficiency of the ith industrialsector in the tth year. Symbols on the right mean the correspondinginfluencing factors in the i th industrial sector in the tth year. b0,b1,.b12 are the unknown coefficients, and εit is the random error.For considering different types of explanatory variables, we

Table

6To

bitregression

resu

ltsof

energy

efficien

cyof

Chineseindustrial

sectors.

Allsample

(I)

(II)

(III)

(IV)

Mod

el1

Mod

el2

Mod

el3

Mod

el4

Mod

el5

Mod

el5

Mod

el5

Mod

el5

Mod

el5

ES0.04

06***(6.10)

0.03

91***(6.25)

0.04

04***(6.14)

0.04

12***(6.62)

0.04

41***(6.27)

�0.017

6**(�

2.28

)0.07

82***(8.16)

0.01

02(0.39)

�0.312

*(�

2.07

)ES

2�0

.000

9***

(�5.27

)�0

.000

8***

(�5.41

)�0

.000

8***

(�5.24

)�0

.000

9***

(�5.69

)�0

.000

9***

(�5.47

)0.00

03**

(2.67)

�0.001

***(�

4.35

)0.00

78***(3.48)

0.01

89(0.76)

IC0.29

09***(4.57)

0.29

65***(4.78)

0.27

5***

(4.16)

0.24

05***(3.23)

0.18

46**

(2.21)

0.18

96(1.28)

0.01

52(0.21)

0.37

04***(2.71)

4.05

09**

(2.69)

PS�0

.471

7***

(�9.22

)�0

.453

9***

(�9.83

)�0

.452

***(�

9.46

)�0

.377

5***

(�6.17

)�0

.340

0***

(�4.48

)�0

.269

8**(�

2.61

)�0

.720

3***

(�5.84

)�0

.344

6***

(�2.99

)�3

.639

3**(�

2.28

)CL

�0.000

6(�

0.96

)�0

.000

8(�

1.11

)0.00

19***(2.81)

�0.004

4(�

1.47

)�0

.01***

(�3.82

)0.00

27(0.68)

MC

�0.058

8(�

0.88

)�0

.039

3(�

0.54

)0.08

54(0.81)

0.49

9**6(2.59

)�0

.126

1(�

1.04

)�0

.781

7(�

1.06

)OC

�0.000

7(�

0.07

)�0

.003

9(�

0.35

)0.55

05(1.42)

2.42

16***(6.14)

�0.001

7(�

0.17

)0.85

22(0.82)

EC0.43

56***(2.71)

0.52

47***(3.88)

0.46

77***(3.41)

0.43

77***(2.49)

0.50

52*(1.95)

1.56

3***

(2.79)

0.65

77**

(1.99)

�0.224

1(�

0.51

)RDI

0.11

09(1.3)

0.09

44(1.11)

0.15

13(0.76)

0.63

***(2.73)

0.05

53(0.72)

1.74

76(1.38)

Ln(RDR)

�0.001

(�0.25

)�0

.000

9(�

0.22

)�0

.004

9(�

1.6)

�0.009

3(�

1.57

)0.00

54(1.06)

�0.039

4*(�

1.75

)Ln

(PC)

�0.001

9(�

0.29

)0.00

13(0.17)

0.03

85**

(2.55)

0.03

33***(3.82)

0.02

22*(1.79)

0.07

52(1.4)

FDI

0.14

22*(1.94)

0.16

23***(2.01)

1.03

23***(3.23)

�0.182

6(�

1.46

)�0

.071

4(�

0.62

)�0

.982

2(�

0.63

)Con

stan

t0.25

7***

(8.50)

0.16

85***(2.91)

0.14

54***(3.11)

0.13

45(1.57)

0.14

00(1.27)

�0.392

5*(�

2.01

)�0

.529

5**(�

2.21

)�0

.047

5(�

0.26

)0.12

94(0.17)

Loglik

elihoo

d11

4.08

121.63

110.57

123.65

113.69

83.70

102.19

78.49

1.98

LRch

i211

8.6***

133.71

***

126.43

***

137.73

***

132.66

***

76.73***

160.79

***

93.50***

17.72

Note:

“*”,“**”,“***”

presenttheirsign

ificance

atleve

lsof

10%,5

%an

d1%

resp

ective

ly.

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e107 105

estimate different Tobit regression models based on the maximumlikelihood estimation, and the corresponding results are presentedin Table 6.

4.2.1. Impacts of enterprise scaleEnterprise scale is one of the important determining factors for

energy utilization efficiency. Estimation results of thewhole samplein Table 6 show that enterprise scale (ES) is overall positively relatedwith energy efficiency, and industry’s energy efficiency will rise byaround 4% when average enterprise scale increases by 1 point. Thismeans that currently, China industry scale has not yet reached theoptimal level. Therefore, increasing the outputs of industries to adesignated scale will contribute to higher scale efficiency andimprove the energy efficiency. However, as the enterprise scale ex-pands, the complexity of industry internal structure leads to moreconsumption on energy and resources, thus may offset the benefitsbrought by scale expansion, decrease the scale efficiency and evenlead to diseconomies of scale. So we take account of quadratic termof enterprise scale (ES2) to measure whether there is a nonlinearrelationship between enterprise scale and energy efficiency. Resultsshow that quadratic term of enterprise scale formed by the wholesample has a negative sign, demonstrating that with the increase ofenterprise scale, industry energy efficiency takes on inverse U-shape, i.e., first rising and then falling. In addition, the estimationresults by group sample show that light industry (II) is in consensuswith that of the whole sample, while enterprise scale of mining in-dustry (I) is positively related with energy efficiency. With expan-sion of enterprise scale, energy efficiency of mining industry (I) andlight industry (III) first drops and then rises, i.e. a characteristic of U-type curve. Hence, in order to meet the demand of economic stra-tegic development and higher energy utilization efficiency, it isneeded to prudently deal with the impacts of enterprise scale onindustrial energy efficiency respectively.

4.2.2. Impacts of industry concentrationThe estimation results of the whole sample in Table 6 show that,

industry concentration is positively related to energy efficiency,implying that higher industry concentration is beneficial to energyefficiency.Analysis onenterprise scale indicates thatenergyefficiencyhas increasing returns to scale.High industryconcentrationmeans anindustrywith largerenterprise scale accordingly, and ithelps improveenergy efficiency. However, group tests indicate that there is no sig-nificant relationship inmining industry (I) and light industry (II). Thisis probably because enterprises with higher degree of industry con-centration are able to attain relative cheaper resources, so they donothave the motivation to improve energy utilization efficiency.

4.2.3. Impacts of property rights structureThe estimation results of the whole sample and the group

sample in Table 6 show that the relationship between propertyrights structure and energy efficiency is significantly positive.Increasing proportion of state-owned enterprises would reduceenergy efficiency, which is in accordance with the current researchconclusions. That mainly lies in the ambiguous property rights andownership scheme as well as the rigid operational mechanism,which hinder the improvement of energy efficiency. Additionally,industries with high pollution are generally capital-intensive, andcapital-intensive industries tend to have a higher state-owned ra-tio. Hence, to further decrease the state-owned ratio and deepenthe reform of state-owned enterprises will benefit industrial en-ergy efficiency.

4.2.4. Impacts of energy consumption structureFrom the estimation results of the whole sample, electricity

consumption is positively related with industrial energy efficiency

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e107106

at the 1% significance level. When electricity consumption pro-portion increases by 1%, industrial energy efficiency will increaseby 43.77%. Estimation results of group tests imply that electricityconsumption is positively related to industrial energy efficiency atthe 10% significance level, and improvements of energy efficiencyin coal, oil and electricity consumptions have a significant positiveimpact on energy efficiency of light industry. In addition, for heavyindustry, coefficients of coal and oil consumptions are not signifi-cant, but negative. This indicates that during manufacturing pro-cess, boosting the proportion of clean energy such as electricity canreduce pollutant emissions like CO2 and enhance the energy effi-ciency of the entire industry.

4.2.5. Impacts of government regulationEstimation results in Table 6 show that investment on pollutant

control has a significant positive impact on upgrading energy effi-ciency. Some studies show that western countries graduallytransfer industries with high energy consumption and high-pollution to China, and this hinders the improvement of indus-trial energy efficiency in China. Moreover, some studies indicatethat China should absorb advanced management experience andtechnologies to best leverage foreign direct investment. In order tohave a better knowledge of the FDI’s impact on Chinese industrialenergy efficiency, we study from two perspectives, i.e. the industryas whole and industrial groups (see Table 6). Estimation results ofthe whole sample demonstrate that FDI benefits the improvementof industrial energy efficiency, while group tests show that theimpacts of FDI on energy efficiency are different between industrycategories. Although some estimated coefficients are not signifi-cant, their signs show that FDI boosts mining industry’s energyefficiency but reduces that of light and heavy industries. Onepossible reason is that, in mining industry FDI could help introduceadvanced manufacturing techniques, management experience andworking process from abroad. Meanwhile, foreign companies makeFDI in China for the affluent and cheap labor force and mineralresources of China, and both light industry and heavy industry arelabor-intensive and resource-intensive. As a consequence, weshould treat FDI differently for different industries and advocatetechnology import so as to enhance the overall industrial energyefficiency in China.

4.2.6. Impacts of other influencing factorsWith regard to labor-capital structure, except formining industry,

increase incapital-labor ratiowillworsenenergyefficiency, especiallyin pollution-intensive industries. Since the reform and opening-up,the capital-labor ratio in China increased and the industry has beenexperiencing a more heavy-oriented trend, which further impedesindustrial energyefficiency from improving.As for technical progress,we use R&D investment and R&D researchers as proxy for techno-logical investment and laborquality respectively. Generally, theybothshould be positively related with energy efficiency; however, theestimated coefficients are not significant. One possible explanation isthediversificationofexpenditureson technological investment, aswecannot peel off the parts relevant to energy efficiency improvement.In addition, in recent years, the government attaches much moreimportance to energy utilization efficiency and high-tech humancapital nurturing, but it will take time for these efforts on energy ef-ficiency improvement to take effect gradually.

5. Conclusion

With the improved Super-SBM model considering undesirableoutputs, in this paper we study the energy efficiency and itsdevelopment tendency of various industrial sectors in China as wellas the four major industrial categories during the period from 2001

to 2010. Furthermore, we compare the energy efficiency differencebetween the “The Tenth Five-year Plan” and “The Eleventh Five-year Plan”, and then explore the influencing factors of energy ef-ficiency. The empirical results show that energy efficiency of eachindustrial sector and category during 2001e2010 has improvedsubstantially. Particularly, during “The Eleventh Five-year Plan”,energy efficiency of various industries has different degrees ofimprovement. However, efficiency variations across various in-dustries have not shown a convergence trend. Energy efficiency oflight industry is the highest among the four categories, followed byheavy industry. But the heavy industry has a higher growth ratecompared with light industry, so the energy efficiency gap betweenthese two industries has been reduced. Energy efficiency variationsindicate prominent transformation in industrial economics.

For the influencing factors of energy efficiency, enterprise scale,industry concentration, industrial property rights structure, andgovernment regulation have significant impacts on energy effi-ciency, and these factors lead to distinctive effects. Enterprise scaleis an important determinant of energy efficiency.With expansion ofenterprise scale, energy efficiency of various industries is reflectedin different U-type curves. Industry concentration is positivelyrelated to energy efficiency, implying that high industry concen-tration benefits the enhancement of energy efficiency. Further re-form of industrial state-owned enterprises will improve industrialenergy efficiency. Moreover, increase in FDI and adjustment inenergy consumption structure will improve industrial energy effi-ciency, especially when electricity consumption ratio is increased.Therefore, through adjusting the property rights structure,strengthening the technological and administrative spillover ef-fects of FDI and developing clear, efficient new energy and substi-tute energy, the industrial energy efficiency will be enhanced to agreat extent. More importantly, the above influencing factors andrelevant policy advice for industrial energy efficiency are raised atthe level of the entire industry as a whole, so the specific industrialgroups and categories need different treatments. We cannot blindlystipulate a uniform energy-saving and emission-cutting target toenhance industrial energy efficiency. We should analyze the attri-butions and development situations of various industries, and dealwith the influencing factors and regulation methods differently.Meanwhile, it is required to pay close attention to the relationshipbetween improvements of energy efficiency and industrial de-velopments, and to adjust this relationship in time.

Acknowledgments

This topic is a stage research outcome of the United States En-ergy Foundation project “Energy subsidy reform and China’s sus-tainable economic development” in 2011 (Project No.: G-1111-15134); Ministry of Education, Philosophy, Social Planning project“Construct renewable energy industry finance risk managementand policy support system based on the life cycle theory” in 2012(Project No.: 12YJAZH056); China Postdoctoral Science Foundation“Sustainable development and social equity: a research based onthe theory of energy subsidies and policy practice” in 2009 (ProjectNo.20090460202); Special Funds for Construction Disciplines inShanxi Province (201205737) and “Returning Students ScientificResearch Project” of Shanxi Province (2012-023). We appreciate thecomments of anonymous reviewers.

References

Andrews-Speed, P., 2009. China’s ongoing energy efficiency drive: origins, progressand prospects. Energy Policy 37, 1331e1344.

Ang, B.W., Liu, F.L., Chung, H.S., 2004. A generalized Fisher index approach to energydecomposition analysis. Energy Econ. 26, 757e763.

H. Li, J.-f. Shi / Journal of Cleaner Production 65 (2014) 97e107 107

Bunse, K., Vodicka, M., Schönsleben, P., Brülhart, M., Ernst, F.O., 2011. Inte-grating energy efficiency performance in production management e gapanalysis between industrial needs and scientific literature. J. Clean. Prod.19, 667e679.

Chen, T.Y., Yeh, T.L., 1998. A study of efficiency evaluation in Taiwan’s banks. Int. J.Ser. Ind. Manag. 9, 402e415.

DeCanio, S.J., 1993. Barriers within firms to energy-efficient investments. EnergyPolicy 21, 906e914.

Färe, R., Grosskopf, S., 2004. Modeling undesirable factors in efficiency evaluation:comment. Eur. J. Oper. Res. 157, 242e245.

Hamilton, J.D., 1983. Oil and the macro economy since WorldWar II. J. Polit. Econ. 91(2), 228e248.

Hasanbeigi, A., Price, L., Lu, H., Lan, W., 2010. Analysis of energy-efficiency oppor-tunities for the cement industry in Shandong Province, China: a case study of 16cement plants. Energy 35, 3461e3473.

Hu, J.L., Wang, S.C., 2006. Total-factor energy efficiency of regions in China. EnergyPolicy 34, 3206e3217.

Hu, Y., 2012. Energy conservation assessment of fixed-asset investment projects: anattempt to improve energy efficiency in China. Energy Policy 43, 327e334.

Jenne, C.A., Cattell, R.K., 1983. Structural change and energy efficiency in industry.Energy Econ. 5, 114e123.

Lardic, S., Mignon, V., 2008. Oil prices and economic activity: an asymmetriccointegration approach. Energy Econ. 30 (3), 847e855.

Oda, J., Akimoto, K., Sano, F., Tomoda, T., 2007. Diffusion of energy efficient tech-nologies and CO2 emission reductions in iron and steel sector. Energy Econ. 29,868e888.

Patterson, M.G., 1996. What is energy efficiency?: concepts, indicators and meth-odological issues. Energy Policy 24, 377e390.

Perc, M., Gómez-Garde�nes, J., Szolnoki, A., Floría, L.M., Moreno, Y., 2013. Evolu-tionary dynamics of group interactions on structured populations: a review. J. R.Soc. Interf. 10, 20120997. http://dx.doi.org/10.1098/rsif.2012.0997.

Perc, M., Szolnoki, A., 2010. Coevolutionary gamesda mini review. BioSystems 99,109e125.

Phylipsen, G., Blok, K., Worrell, E., 1997. International comparisons of energyefficiency methodologies for the manufacturing industry. Energy Policy 25,715e725.

Ray, S.C., 2004. Data Envelopment Analysis: Theory and Techniques for Economicsand Operations Research. Cambridge University Press.

Seiford, L.M., Zhu, J., 2002. Modeling undesirable factors in efficiency evaluation.Eur. J. Oper. Res. 142, 16e20.

Song, M.L., Wu, J., Wang, Y.M., 2012. Reducing energy demand in China: a statisticalanalysis of urban energy consumption in Anhui Province. Energy Environ. 23,17e32.

Song, M.L., Zhang, L.L., An, Q.X., Wang, Z.Y., Li, Z., 2013. Statistical analysis andcombination forecasting of environmental efficiency and its influential factorssince China entered the WTO: 2002e2010e2012. J. Clean. Prod. 42, 42e51.

Sun, J.W., 1998. Changes in energy consumption and energy intensity: a completedecomposition model. Energy Econ. 20, 85e100.

Tone, K., 2001. A slacks-based measure of efficiency in data envelopment analysis.Eur. J. Oper. Res. 130, 498e509.

Tone, K., 2002. A slacks-based measure of super-efficiency in data envelopmentanalysis. Eur. J. Oper. Res. 143, 32e41.

Tone, K., 2003. Dealing with Undesirable Outputs in DEA: a Slacks-based Measure(SBM) Approach. In: GRIPS Research Report Series, pp. 1e5.

Tone, K., Sahoo, B.K., 2003. Scale, indivisibilities and production function in dataenvelopment analysis. Int. J. Prod. Econ. 84, 165e192.

Wang, Z.H., Zeng, H.L., Wei, Y.M., Zhang, Y.X., 2012. Regional total factor energy effi-ciency: an empirical analysis of industrial sector in China. Appl. Energy97,115e123.

Yang,W., Han, A.,Wang, S.Y., 2013. Analysis of the interaction between crude oil priceand US stock market based on interval data. Int. J. Energy Stat. 1 (2), 85e98.

Zhou, N., Levine, M.D., Price, L., 2010. Overview of current energy-efficiency policiesin China. Energy Policy 38, 6439e6452.